abliteratedmodels.org

Muse-Glimmer-Preserving-Abliterated

Muse-Glimmer-Preserving-Abliterated is a refusal-ablated variant of Blackroot/Muse-Glimmer-30GB-Preserving-Abliterated, published on Hugging Face by Blackroot. It is apache-2.0 licence. It is distributed across 4 repositories in Transformers, GGUF and GGUF (imatrix) formats, totalling 6K downloads.

Abliteration (directional ablation)Other6,0432

Specification

Model name
Muse-Glimmer-Preserving-Abliterated
Publisher
Blackroot
Parameters
Not determinable from name
Hardware class
Unknown
Licence
apache-2.0
Task
Image text to text
Context window
Not documented
Ablation scope
Not stated by the publisher
Formats available
Transformers, GGUF, GGUF (imatrix)
First indexed
15 Aug 2026
Last updated
18 Aug 2026

Ablation technique

How Muse-Glimmer-Preserving-Abliterated was modified, and what that implies.

Abliteration (directional ablation)

The single residual-stream direction that mediates refusal is identified from harmful/harmless prompt pairs and projected out of the model weights.

Reported scope: Not stated by the publisher. Publishers frequently omit this, so verify behaviour empirically rather than assuming full coverage.

Running it

Commands are templates — confirm the exact repository and quant file before use.

llama.cpp / GGUF
hf download mradermacher/Muse-Glimmer-30GB-Preserving-Abliterated-i1-GGUF --local-dir ./muse-glimmer-preserving-abliterated
llama-cli -m ./muse-glimmer-preserving-abliterated/<file>.gguf -p "..."
Transformers
from transformers import AutoModelForCausalLM, AutoTokenizer

repo = "Blackroot/Muse-Glimmer-30GB-Preserving-Abliterated"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
    repo, torch_dtype="auto", device_map="auto"
)
vLLM
vllm serve Blackroot/Muse-Glimmer-30GB-Preserving-Abliterated --trust-remote-code

Downloads and variants (3)

Every published repository of Muse-Glimmer-Preserving-Abliterated, including quantised re-releases by other authors.

Security-team assessment

Derived from this model’s metadata and the publisher’s claims — not from benchmarks run by this site.

  • Muse-Glimmer-Preserving-Abliterated will attempt offensive-security prompts that a hosted commercial model declines, which is what makes it usable for red-team corpus generation and for measuring what an unaligned model of this class produces.
  • The publisher does not state which layers were ablated, so assume nothing about refusal consistency — probe it directly.
  • Declared licence is apache-2.0, but the terms that bind you are Blackroot/Muse-Glimmer-30GB-Preserving-Abliterated’s — a re-release cannot grant more than its parent.

Handling: run it isolated, put moderation in front of it if anyone outside your team can reach it, and verify the weights before loading — these are third-party artifacts. Full handling and licence guidance.

Metadata for Muse-Glimmer-Preserving-Abliterated is collected from the public Hugging Face API and the publisher’s model card. This site does not host weights, is not affiliated with Blackroot, and does not independently verify publisher claims. See responsible use.